Comparison of prediction of biogas yield in a batch mode underground fixed dome digester with cow dung

Abstract

The study aimed to compare the development of an artificial neural network (ANN) and multilinear regression (MLR) model used to predict the performance of biogas in a batch-mode underground fixed dome biogas digester. In this study, 50 experimental datasets were used to assess the rate of biogas production with developed ANN and MLR models. The six variables, including solar irradiance, relative humidity, slurry temperature, biogas temperature, pH, and ambient temperature, were selected as the input parameters or predictors of the model. Therefore, the developed ANN and MLR models were used to describe the rate of biogas yield. The study found that the determination coefficient (R2) and root mean square error (RMSE) for ANN and MLR were 0.999/0.968 and 8.33 × 10−6/1.84 × 10−4, respectively. Both models were significant because of their high correlation between measured and predicted values of the biogas yield. However, the ANN performs better because of the smaller RMSE and higher R2 derived compared to the corresponding values of the MLR. The study proved that both the ANN and MLR can accurately predict the rate of biogas production but with better predictions obtained from ANN.